Wikipedia:Random pages test

This visualisation of Wikipedia's coverage is based on a random pages test (User:Smallbones/1000 random results).

Suppose you want to answer some broad question about Wikipedia. Is it any good? What subjects does it cover? In what proportions? Is there much vandalism?

Finding a truly accurate answer would mean reading every article. However, Wikipedia has quite a few articles. Studying a random selection is a more practical approach to get a grasp on these questions – and compiling this sample is as simple as hitting Special:Random a bunch of times to record what comes up.

Random pages tests by various editors can be found in Category:Random pages tests, although the category is not comprehensive. The concept of random sampling is not exactly original to Wikipedia; indeed, various editors seem to have independently conceived the idea in Wikipedia's context several times over. The earliest tests date back to 2003. The number of articles in these tests ranges from as few as ten to as many as one thousand.

Selected examples

edit
Selected examples of random pages tests
Link Date Sample User Notes
meta:English Wikipedia Quality Survey October 2003 200 Adam Carr An early example. Possibly the earliest.
User:Matt Crypto/RandomArticles September 2004 100 Matt Crypto Employed a script to weight articles by pageviews. See also User:Stormie/Random.
User:Carnildo/The 100 November 2005 100 Carnildo A longitudinal study – sample followed up three times in 2006, then again in 2015.
User:Opabinia regalis/Article statistics November 2006 250 Opabinia regalis A large sample for the time.
Wikipedia:Counter-Vandalism Unit/Vandalism studies/Study1 February 2007 174 Collaboration Focus on vandalism prevalence.
User:Knulclunk/Random September 2008 200 Knulclunk Visualised with pie charts.
Wikipedia:Wikipedia Signpost/2009-06-22/Vandalism June 2009 100 Loren Cobb (Aetheling) Focus on vandalism prevalence.
User:Smallbones/1000 random results December 2015 1000 Smallbones A large sample.
User:North8000/Display 2021 1000 North8000 Another large sample.